DLR-RM / rl-baselines3-zoo

A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.
https://rl-baselines3-zoo.readthedocs.io
MIT License
1.9k stars 495 forks source link

[Question] Should "continue training" load the vecnormalize.pkl as well as the model.zip? #386

Closed xibeisiber closed 1 year ago

xibeisiber commented 1 year ago

❓ Question

Hi, I saw in zoo3 "continue training" will load the pretrained model.zip at the beginning.

I wonder should the corresponding vecnormalize.pkl file be loaded as well?

Thanks!

Checklist

araffin commented 1 year ago

Hello, it should and it is: https://github.com/DLR-RM/rl-baselines3-zoo/blob/2d5db11f6aee6d231d2154849715b0f7a9435038/rl_zoo3/exp_manager.py#L553-L563

In the worst case, statistics tends to quickly converge if started from scratch.

xibeisiber commented 1 year ago

Thanks!